Autonomous trading agent running on Base chain. Cycles on a schedule, scores market signals, analyzes on-chain activity, and makes decisions without human intervention. Currently in paper trading mode.
An autonomous agent that runs on a fixed schedule, gathers market data from multiple sources, scores trading signals, reads on-chain activity, and decides whether to buy, sell, or do nothing. It runs continuously as a systemd service. No human in the loop. No manual approvals. Just an agent making decisions.
Autonomous agents that move real money need to prove themselves first. The agent has completed over 1,200 cycles with 11 real paper trades. The results are break-even. More importantly, the architecture has held. No blow-ups. No runaway positions. No unbounded losses. The safety mechanisms work as designed. That is the real win.
Paper mode means the agent goes through the full decision pipeline and records what it would have done. The data builds confidence before real capital enters.
Not a high-frequency trading bot. Not a meme coin gambler. Not something you should copy and run with real money. It is an architecture experiment that happens to trade crypto. The interesting part is the autonomous decision-making pipeline, not the asset class.
Built with Python, Web3.py, Ollama, and SQLite. Runs on a local machine. Paper mode only. Not open source.